Telling arriving demand from produced turnover

The short answer

Arriving demand and produced turnover share exactly one property: the headline figure. They differ in how many participants are behind it, how those participants were funded, how the trades are spaced and sized, what everyone holds at the end, and how sharply the activity starts and stops.

None of those differences is conclusive on its own. Every one of them has an ordinary explanation that fits it just as well, which is why a reading is assembled from several tells whose ordinary explanations contradict each other, rather than from whichever tell looked most striking.

What the two terms mean here

Operational definitions work better than moral ones. Call turnover demand-driven when the decisions behind it are independent, meaning one participant trading does not mechanically cause another to trade. Call it produced when one decision-maker is behind trades that are presented as if they came from separate participants.

Notice what those definitions leave out. They say nothing about honesty, profit or legitimacy. A market maker quoting both sides is one decision-maker behind many trades and is a useful participant. A crowd reacting to the same post is many decision-makers whose choices are anything but independent. Both definitions have awkward edges, and every tell below inherits them.

The word produced is used throughout rather than fake or artificial. Fake implies the trades did not happen, which is false: they settled, they paid fees, they are permanently recorded. Artificial carries a judgement the evidence cannot support. Produced describes the mechanism and nothing more, which is the most a transaction record can justify.

Twelve tells and their ordinary explanations

Twelve fields do most of the separating work. Each is stated with the ordinary explanation that defeats it, because a tell without its counter-argument is decoration rather than evidence.

Read that last column carefully, because it is the part that matters. Twelve tells with twelve defeating explanations is not twelve pieces of evidence. It is twelve invitations to be wrong, and the only way to use them well is to look at which explanations can be true simultaneously.

Why incompatibility beats counting

The instinct is to count tells and treat six as better than three. That instinct is wrong, and understanding why is the single most useful idea in this note.

Suppose a window shows regular spacing, a narrow size band and flat end inventory. Three tells. Now notice that one ordinary explanation covers all three at once: a scheduled rebalancing process would produce regular spacing because it is scheduled, a narrow size band because it trades fixed amounts, and flat inventory because rebalancing returns to a target. Three tells, one explanation, and therefore one observation counted three times.

Now suppose a different window shows sustained address novelty alongside funding paths converging on a small number of sources. Two tells. But the ordinary explanation for the first, that genuinely new people are arriving, is not compatible with the ordinary explanation for the second, that a common funding source is coincidental. Two tells that cannot both be explained away are worth more than six that can.

Reasoning without a base rate

The obvious next step is to build a score. Assign weights to the twelve tells, sum them, produce a number. This desk does not do that, and the reason is worth setting out because the omission looks like laziness and is the opposite.

A score that means anything requires knowing how often produced windows and demand-driven windows each produce each tell. Those are base rates, and this desk does not have them. Reading windows that came to attention because they looked unusual cannot produce an unbiased sample, venue coverage is rarely complete, and no labelling process here has a measured error rate. Weights invented without those inputs would be arbitrary numbers wearing a rigorous costume.

What can be done instead is narrower and honest. Produce a shortlist of surviving explanations, state which tells eliminated the others, and assign a confidence band by rule rather than by arithmetic.

1. List the explanations that fit the windowWrite them out before measuring, including the boring ones. Explanations added after the data has been seen tend to arrive only when they support the reading already forming.
2. Cross off the ones a structural fact eliminatesReserve changes and pool creation are recorded, so they remove explanations decisively rather than by argument.
3. For each surviving explanation, name the tells it cannot account forThis is the step that does the work, and it is uncomfortable, which is why it has to be written down rather than done in your head.
4. Keep every explanation that survivesIf two survive, the reading names two. Forcing a choice between surviving explanations is how a desk drifts towards whichever conclusion is more quotable.
5. Set the band by what remainsOne explanation surviving with several incompatible tells behind it is Working. One explanation surviving on a single tell is Provisional. Two or more surviving with nothing to separate them is Undecidable.

That procedure yields a shortlist and a band rather than a number. It is less satisfying to publish and much harder to misquote, which is roughly the correct trade for this subject.

When both are present at once

The framing of demand against production suggests a window is one or the other. Frequently it is both, and the mixed case is the one most likely to be misread in either direction.

A pair can carry produced activity as a floor while genuine participants arrive on top of it, and the aggregate distribution will then sit between the two profiles: partially quantised sizes, partly regular spacing, an address list that grows but not as fast as a pure demand window would. Read as a whole, that looks ambiguous. Read per participant group, it often separates cleanly.

The practical move is to stop treating the window as one population. Split the participants by funding origin or by size profile, then measure the tells within each group separately. A group showing tight spacing and flat inventory sitting alongside a group showing wide dispersion and directional inventory is a much more specific finding than either an ambiguous aggregate or a forced single label.

The supply side, plainly described

Produced turnover exists because it can be bought, and describing that plainly improves readings rather than compromising them. A SOL volume bot is software that spreads activity across funded wallets according to settings an operator chooses, and it is sold, priced and documented like any other tool.

Three things follow directly. The parameter set is small, so the output is shaped by a handful of fields rather than by anything elaborate. Defaults exist and are frequently left alone, so unrelated runs share features. And the whole thing runs against a budget, which produces the sharp boundaries that turn out to be the most durable tell in the table above.

Motives on the production side vary enormously and none of them is visible on chain. Some activity exists to meet a listing threshold, some to keep a pair from looking abandoned, some to support a market maker on a thin book. Those are very different situations, and treating them as one category is the mirror image of the mistake this note is trying to prevent.

Writing the conclusion

The output of this comparison is a paragraph with required parts. Anything missing turns a measurement into an assertion, and assertions about markets get repeated far past the evidence that started them.

• The window and the comparison period, both fixed before measurement.
• Venue coverage, including what was knowingly excluded.
• Which of the twelve tells were measured, including those that returned nothing.
• The ordinary explanation considered for each tell that pointed somewhere.
• Whether those explanations were mutually compatible, stated explicitly.
• The surviving explanations, all of them, with no hedging option added.
• The confidence band and why it is not the band above it.
• The single observation that would overturn the reading.
• A sentence naming what the reading does not claim, including intent.

Anonymised labels are used throughout wherever intent has not been established, which in practice means always. A reading that describes a group as an operator has already assumed the thing it was meant to test.

Where this stops working

All twelve tells degrade on short windows, thin pairs and incomplete venue coverage, and they degrade quietly rather than obviously. A distribution measured on forty trades will look like whatever the first forty traders happened to do, and it will still produce a confident-looking chart.

They also degrade against effort in a way worth naming plainly. Every tell in the table describes a default rather than a limit, and an operator who cared about any particular tell could address it. This note does not describe how, and the analytical consequence is simply that the absence of a tell is very weak evidence. Finding a tell says something; failing to find one says almost nothing, and readings should be written asymmetrically to reflect that.

None of them reaches intent, and no combination of them does either. A window whose profile matches the produced column has been described, not accused. Whether anyone involved considered the activity legitimate is a question the record does not contain, and this desk does not answer it about any named party.

Finally, the desk does not publish evasion guidance. Where this note naturally raises how a footprint could be reshaped to defeat a tell, it stops there. Explaining how a check works helps a reader interpret what they are looking at; explaining how to defeat it helps only somebody with a different aim. The line between the two is easy to hold in practice, and holding it costs this note nothing a reader needed.

Questions the desk gets asked